Noise suppression and component extraction of underwater acoustic signals

Nonstationary characteristics are present in some underwater acoustic signals. In this paper, the authors perform noise reduction and isolation of nonstationary acoustic signal components using time-frequency filtering. Furthermore, they discuss the necessary mathematical conditions for achieving the best possible reconstruction during the synthesis part of a filtered Short-Time Fourier Transform. They then outline the steps for designing the filters using noise statistics and morphological operations. Finally, they offer examples where noise resulting from the interaction between the hydrophone and the water mass is reduced and components are isolated using the proposed approach.